File size: 3,994 Bytes
c24037e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 |
This is the repo of **PointOBB-v2: Towards Simpler, Faster, and Stronger Single Point Supervised Oriented Object Detection**
![pipeline]()
### Installation
Please refer to the [Installation](https://github.com/open-mmlab/mmrotate/blob/main/README.md), we copy it here.
```
conda create -n open-mmlab python=3.7 pytorch==1.7.0 cudatoolkit=10.1 torchvision -c pytorch -y
conda activate open-mmlab
pip install openmim
mim install mmcv-full
mim install mmdet
git clone https://github.com/taugeren/PointOBB-v2.git
cd mmrotate
pip install -r requirements/build.txt
pip install -v -e .
```
### Data Preparation
Please follow [data_preparation](https://github.com/taugeren/PointOBB-v2/tree/main/tools/data) to prepare formatting data
### Train CPM
If you want to visualize CPM result during training, please set **visualize=True** in **train_config**
Please modified the config code that contains the **visualize directory path**
The learning rate for **n GPU card** and **batch size m** is 0.0125 * n * m
For single GPU
```
# Basic format: python tools/train.py ${CONFIG_FILE} [optional arguments]
python tools/train.py configs/pointobbv2/train_cpm_dotav10.py --work-dir work_dirs/cpm_dotav10 --gpu-ids 0
```
For multiple GPU
```
# Basic format: ./tools/dist_train.sh ${CONFIG_FILE} ${GPU_NUM} [optional arguments]
CUDA_VISIBLE_DEVICES=0,1 PORT=29801 ./tools/dist_train.sh configs/pointobbv2/train_cpm_dotav10.py 2
```
### Generate Pseudo Label
Please modified the config code that contains the **directory path**
For single GPU
```
# Basic fromat: python tools/train.py ${CONFIG_FILE} --resume-from ${CPM_CHECKPOINT_FILE} [other arguments]
python tools/train.py configs/pointobbv2/generate_pseudo_label_dotav10.py --resume-from work_dirs/cpm_dotav10/epoch_6.pth --work-dir work_dirs/cpm_dotav10 --gpu-ids 0
```
For multiple GPU
```
# Basic format: ./tools/dist_train_resume.sh ${CONFIG_FILE} ${CPM_CHECKPOINT_FILE} ${GPU_NUM} [optional arguments]
CUDA_VISIBLE_DEVICES=0,1 PORT=29801 ./tools/dist_train_resume.sh /ssd1/renbotao/github_submission/mmrotate/configs/pointobbv2/generate_pseudo_label_dotav10.py work_dirs/cpm_dotav10/epoch_6.pth 2
```
### Train Detector
You can use different oriented object detection detector in [MMRotate Config](https://github.com/open-mmlab/mmrotate/tree/main/configs)
Please modify the **pseudo label path** in config file
For example, using Redet, the training command:
```
# single GPU
python tools/train.py configs/pointobbv2/redet_dotav10.py --work-dir work_dirs/cpm_dotav10 --gpu-ids 0
# multiple GPU
CUDA_VISIBLE_DEVICES=0,1 PORT=29801 ./tools/dist_train.sh configs/pointobbv2/redet_dotav10.py 2
```
the testing command:
```
# single GPU
python tools/test.py work_dirs/redet_dotav10/redet_dotav10.py work_dirs/redet_dotav10/epoch_12.pth --gpu-ids 0 --format-only --eval-options submission_dir=testmodel/redet_dotav10_epoch12
# multiple GPU
CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=29816 tools/dist_test1.sh work_dirs/redet_dotav10/redet_dotav10.py work_dirs/redet_dotav10/epoch_12.pth 4
```
### Our log and checkpoint
| Dataset | Config | Log | Checkpoint |
| :-------: | :----------------------------------------------------------: | :----------------------------------------------------------: | :---------------------------: |
| DOTA-v1.0 | [config](https://github.com/taugeren/PointOBB-v2/tree/main/configs/pointobbv2) | [quark](https://pan.quark.cn/s/72c9fb568db8) [hugging face]() | [quark]() [hugging face]() |
| DOTA-v1.5 | [config](https://github.com/taugeren/PointOBB-v2/tree/main/configs/pointobbv2) | [quark]() [hugging face]() | [quark]() [hugging face]() |
| DOTA-v2.0 | [config](https://github.com/taugeren/PointOBB-v2/tree/main/configs/pointobbv2) | [quark]() [hugging face]() | [quark]() [hugging face]() |
|